Re: possible deadlock in start_this_handle (2)

[Tetsuo Handa]

On 2021/02/15 21:45, Jan Kara wrote:
> On Sat 13-02-21 23:26:37, Tetsuo Handa wrote:
>> Excuse me, but it seems to me that nothing prevents
>> ext4_xattr_set_handle() from reaching ext4_xattr_inode_lookup_create()
>> without memalloc_nofs_save() when hitting ext4_get_nojournal() path.
>> Will you explain when ext4_get_nojournal() path is executed?
> 
> That's a good question but sadly I don't think that's it.
> ext4_get_nojournal() is called when the filesystem is created without a
> journal. In that case we also don't acquire jbd2_handle lockdep map. In the
> syzbot report we can see:

Since syzbot can test filesystem images, syzbot might have tested a filesystem
image created both with and without journal within this boot.

> 
> kswapd0/2246 is trying to acquire lock:
> ffff888041a988e0 (jbd2_handle){++++}-{0:0}, at: start_this_handle+0xf81/0x1380 fs/jbd2/transaction.c:444
> 
> but task is already holding lock:
> ffffffff8be892c0 (fs_reclaim){+.+.}-{0:0}, at: __fs_reclaim_acquire+0x0/0x30 mm/page_alloc.c:5195
> 
> So this filesystem has very clearly been created with a journal. Also the
> journal lockdep tracking machinery uses:

While locks held by kswapd0/2246 are fs_reclaim, shrinker_rwsem, &type->s_umount_key#38
and jbd2_handle, isn't the dependency lockdep considers problematic is

  Chain exists of:
    jbd2_handle --> &ei->xattr_sem --> fs_reclaim

   Possible unsafe locking scenario:

         CPU0                    CPU1
         ----                    ----
    lock(fs_reclaim);
                                 lock(&ei->xattr_sem);
                                 lock(fs_reclaim);
    lock(jbd2_handle);

where CPU0 is kswapd/2246 and CPU1 is the case of ext4_get_nojournal() path?
If someone has taken jbd2_handle and &ei->xattr_sem in this order, isn't this
dependency true?

> 
> rwsem_acquire_read(&journal->j_trans_commit_map, 0, 0, _THIS_IP_);
> 
> so a lockdep key is per-filesystem. Thus it is not possible that lockdep
> would combine lock dependencies from two different filesystems.
> 
> But I guess we could narrow the search for this problem by adding WARN_ONs
> to ext4_xattr_set_handle() and ext4_xattr_inode_lookup_create() like:
> 
> WARN_ON(ext4_handle_valid(handle) && !(current->flags & PF_MEMALLOC_NOFS));
> 
> It would narrow down a place in which PF_MEMALLOC_NOFS flag isn't set
> properly... At least that seems like the most plausible way forward to me.

You can use CONFIG_DEBUG_AID_FOR_SYZBOT for adding such WARN_ONs on linux-next.